lncRScan-SVM: A Tool for Predicting Long Non-Coding RNAs Using Support Vector Machine

Abstract

Functional long non-coding RNAs (lncRNAs) have been bringing novel insight into biological study, however it is still not trivial to accurately distinguish the lncRNA transcripts (LNCTs) from the protein coding ones (PCTs). As various information and data about lncRNAs are preserved by previous studies, it is appealing to develop novel methods to identify the lncRNAs more accurately. Our method lncRScan-SVM aims at classifying PCTs and LNCTs using support vector machine (SVM). The gold-standard datasets for lncRScan-SVM model training, lncRNA prediction and method comparison were constructed according to the GENCODE gene annotations of human and mouse respectively. By integrating features derived from gene structure, transcript sequence, potential codon sequence and conservation, lncRScan-SVM outperforms other approaches, which is evaluated by several criteria such as sensitivity, specificity, accuracy, Matthews correlation coefficient (MCC) and area under curve (AUC). In addition, several known human lncRNA datasets were assessed using lncRScan-SVM. LncRScan-SVM is an efficient tool for predicting the lncRNAs, and it is quite useful for current lncRNA study.

Long Noncoding RNA Identification: Comparing Machine Learning Based Tools for Long Noncoding Transcripts Discrimination

“…LncRScan-SVM [33] classifies the sequences mainly by evaluating the qualities of nucleotide sequences, codon sequence, and transcripts structure. The counts and average length of exon in one sequence are calculated.…”

“…For 10,000 protein-coding genes and 10,000 lncRNAs selected from UCSC genome browser (GRCh37/hg19), CPC picked up about 97.62% coding transcripts while CPAT distinguished 85.28% of them. CPAT also outperforms CPC with 89.94% accuracy [33]. Table 3 shows the performance of these tools on the same testing dataset.…”

“…For sensitivity, lncRScan-SVM obtained 93.88% and 95.29% on the same testing datasets, which are only 0.72% and around 0.1% lower than CPAT's results, respectively, but much higher than CPC's 67.23% and 75.46%. In addition, lncRScan-SVM also has the best results of accuracy and AUC [33] on these datasets.…”

“…Here, comprehensive analyses and discussions of these tools were provided. Then, we compared several popular machine learning based methods, including Coding Potential Calculator (CPC) [28], Coding Potential Assessment Tool (CPAT) [29], Coding-Non-Coding Index (CNCI) [30], predictor of long noncoding RNAs and messenger RNAs based on an improved k -mer scheme (PLEK) [31], Long noncoding RNA IDentification (LncRNA-ID) [32], and lncRScan-SVM [33]. In addition, lncRNA-MFDL [34] and LncRNApred [35], two artificial neural network- (ANN-) involved tools, are also introduced in this paper.…”

Transcriptome Analysis for Non-Model Organism: Current Status and Best-Practices

“…Accumulating evidence showed that the protein-coding genes are accounted for only 50% of inal assembled transcriptome data. Mining inal non-redundant transcriptome data via long non-coding RNA identiication tools such as PLEK [90], lncRScan-SVM [91], FEELnc [92] or measuring protein coding potential of transcripts using various tools such as coding potential calculator (CPC) [93], coding potential assessment tool (CPAT) [94], coding-non-coding index (CNCI) [95] provides us more information about the transcriptome landscape of non-model organism.…”